mcp-client-server-example

mcp-client-server-example

1
ai_chatbotdeveloper_tools

This project illustrates using an AI agent that interprets and executes user commands locally through Python functions, using MCP and Ollama. It promises autonomy without manual coding, operating fully offline to enable smart, secure AI workflows.

MCP + Ollama Local Tool Calling Example

This project demonstrates how a local AI agent can understand user queries and automatically call Python functions using:

  • Model Context Protocol (MCP)
  • Ollama for running a local LLM (e.g., Llama3)
  • Python MCP Client and Server

šŸ”— Sequence Diagram

sequenceDiagram
    participant User
    participant MCP_Client
    participant Ollama_LLM
    participant MCP_Server

    User->>MCP_Client: 1) User types: "What is 5 + 8?"
    MCP_Client->>Ollama_LLM: 2) Send available tools + user query
    Ollama_LLM->>Ollama_LLM: 3) Understand query & tool descriptions
    Ollama_LLM->>Ollama_LLM: 4) Select tool: add(a=5, b=8)
    Ollama_LLM->>MCP_Client: 5) Return tool_call
    MCP_Client->>MCP_Server: 6) Execute add(a=5, b=8)
    MCP_Server-->>MCP_Client: 7) Return result: 13
    MCP_Client-->>User: 8) Show final answer: 13

šŸ“š Project Structure

.
ā”œā”€ā”€ math_server.py      # MCP Server exposing add() and multiply() tools
ā”œā”€ā”€ ollama_client.py    # MCP Client interacting with Ollama
ā”œā”€ā”€ README.md           # Project documentation

šŸ› ļø Setup Instructions

1. Install Requirements

pip install "mcp[cli] @ git+https://github.com/awslabs/mcp.git" openai==0.28 httpx

Make sure you have Ollama installed and running.

2. Pull or run an LLM model

ollama run llama3

(Ensure the model you run supports tool calling.)

3. Run the MCP Server

python math_server.py

The server exposes two simple tools:

  • add(a: int, b: int) -> int
  • multiply(a: int, b: int) -> int

4. Run the MCP Client

python ollama_client.py math_server.py

5. Interact!

Example queries:

Query: What is 5 + 8?
Response: 13

Query: Multiply 7 and 9
Response: 63

The MCP client sends the query and available tools to Ollama. The LLM internally decides which tool to use based on the tool descriptions and user intent.

šŸš€ How It Works

  • MCP Client lists available tools.
  • Sends tools + user query to Ollama LLM.
  • LLM reasons about the best matching tool.
  • LLM generates a tool_call.
  • MCP Client invokes the function via the MCP Server.
  • Final result is returned and displayed.

āœ… No manual hardcoding! āœ… Everything runs locally! āœ… Fully autonomous!

šŸ“¢ Why This Matters

This pattern enables building smart local AI agents that:

  • Understand user intent
  • Dynamically select the correct actions
  • Operate fully offline and locally

It opens doors for:

  • Autonomous developers
  • Local intelligent assistants
  • Secure AI workflows

šŸ·ļø Hashtags for Sharing

#MCP #ModelContextProtocol #Ollama #LocalLLM #FunctionCalling #Python #AI #DeveloperTools #AIEngineering #AutonomousAgents

šŸ™Œ Credits

"Smarter AI agents start with understanding how they think!"

Next Steps: Add Streamlit UI or Dockerize this project šŸš€